Virtual New Nitrogen, Phosphorus, Water Input, and Greenhouse Gas Emission Indicators for the Potatoes Consumed in China

Abstract

Based on China’s potato staple food policy, we chose potatoes as a study case to analyze the following indicators—the virtual new nitrogen factor (VNNF), the virtual new phosphorus factor (VNPF), the virtual irrigation-water (IW) factor (VIWF), the virtual total water (IW + precipitation) factor (VTWF), and the virtual greenhouse gas (GHG) emission factor (VCF) of the potatoes consumed by households in the six potato cultivation regions—by reviewing 220 publications from 2000 to 2020. The results showed that the VNNF ranged between 17.8 ± 7.8 and 30.1 ± 17.0 kg N kg⁻¹ N in the consumed potato, the VNPF ranged between 8.4 ± 5.0 and 18.8 ± 11.3 kg P kg⁻¹ P in the consumed potato, the VIWF ranged between 0.3 ± 1.0 and 1.8 ± 1.4 m³ IW kg⁻¹ for the consumed standard yield (except in the three mainly rainfed potato regions), the VTWF ranged between 4.8 ± 2.2 and 9.3 ± 3.7 m³ total water kg⁻¹ for the consumed standard yield, and the VCF ranged between 3.4 ± 1.5 and 5.9 ± 2.4 kg CO₂ equivalent kg⁻¹ for the consumed standard yield, under the conventional practice in the six potato cultivation regions. The normalization score results indicate that the northeast, northwest, southwest, and south of China are relatively more suitable regions in which to plant potatoes, based on the VNNF, VNPF, VIWF or VTWF, and VCF indicators.

1. Introduction

With the increasing frequency of food trade between different regions, the virtual resource inputs—nitrogen (N), phosphorus (P), water, and land, and the environmental footprint indicators—N, P, and carbon dioxide (CO₂) (hereafter defined as the resource inputs or environmental footprints for produce per unit of the final resource consumed in foods or per unit of the final consumed food product in the field production stages) have been used to track the actual resource inputs and environmental effects of the food consumption by households of a given food item (e.g., personal, city, province, and nation) [1,2,3,4]. The virtual resource inputs and environmental footprints of the imported or exported food can be calculated by the actual amount of the imported or exported food multiplied by the corresponding parameters of the virtual N, P, and water resource inputs or the environmental costs per unit of the food items in their production process [5,6,7]. In recent years, the Chinese government has proposed “The 14th Five-Year Plan for National Agricultural Green Development (2021)” and “The 14th Five-Year Plan for National Livestock and Poultry Manure Utilization, Combination of Farming and Animal Husbandry Construction Plan (2021).” These plans aim to reduce greenhouse gas (GHG) emissions, ammonia emissions, nutrient leaching, and agricultural waste generation by optimizing agricultural crop practices at the field level. The study of the virtual resource inputs and GHG emissions of the food trade could help to further determine the key influencing factors of the resource inputs and environmental costs of different foods from harvest to table, in addition to field management practices, and give some guidance for reducing the resource inputs and environmental costs of the food trade in the entire field production–supply–consumption chain.

To track the actual resource inputs and environmental effects of the final food consumption from the food production areas, the virtual N factor (VNF) of foods has been proposed, which is defined as the initial new N—including the N from chemical fertilizers, the N in irrigation water, the atmospheric N deposition, and the biological N₂ fixation (BNF)—input per unit of food in the production process, which does not include the N in the food itself [1,8,9]. Some studies have reported the VNFs for different food items in China (e.g., grains, tubers, beans, vegetables, fruits, and various meat and animal products) [10,11]. Nevertheless, the current VNFs of different foods underestimate the virtual N input to the food system for the production per unit of available N in the consumed food items at their production level. This is mainly due to the VNF’s failure to calculate the commodity ratio (defined as the proportion of the produced food, which has the appropriate size and quality to be consumed, to the total yield [12,13,14]) and the ratio between a household’s available food item supplies and the actual consumption [14,15]. The ratio of a household’s available food supply to its actual consumption in China varied from 1.3 to 4.3 for various food items from 1990 to 2015 [15]. The water use efficiency (WUE) indicator is defined as the kg of grain per m³ of water applied or the water consumption per unit of grain [16], and the GHG intensity (GHGI) is defined in kilograms of GHG emissions per unit of product yield [17,18,19]. The WUE and GHGI indicators have been widely adopted to evaluate and compare the water efficiency and GHG cost of different crops at the field level [15,19,20]. Normally, the WUE and GHGI indicators are directly used to estimate the total water resource input and the total GHG emissions in the area where the food is produced [3]. These two indicators are also unsuitable for the same reasons as the VNF indicator to calculate the total water input and GHG emissions in the food exporting areas, which are also driven by the food consumption of households from a consumption perspective. In addition, several studies have reported that the N, P, and water resource inputs and GHG emissions for different food items have large regional variations in their cultivation stages [3,20] because China has multiple crops each year in a broad range of soil–climatic regimes; moreover, large changes in the technological level and production mode, lead to significant differences in resource inputs and yields [20,21]. Notably, the regional-scale virtual resource inputs and environmental footprint indicators of the consumed food items are of great importance to the results of a multiregional input–output model simulation. Moreover, the regional-scale virtual resource inputs or environmental cost indicators of the different consumed foods can also help to improve the resource use efficiency and decrease the GHG emissions in a given food system by seeking the relatively highly efficient production regions from the food supply chain perspective [22,23].

In this study, we propose the virtual new N and P factors (VNNF and VNPF) concept, which is defined as the initial new N and P inputs in the production stages per unit of the actually consumed food N or P. In addition, we introduce the virtual irrigation-water or total water (referred to hereafter as irrigation water + precipitation) factor (VIWF and VTWF, defined as the irrigation water or total water amount per unit of the actually consumed food) and the virtual CO₂ equivalent (CO₂e) emission factor (VCF, defined as the net GHG emissions in the production area per unit of the foods actually consumed). We aim to use these indicators to track the virtual new N and P inputs, irrigation water and total water consumption, and the virtual GHG emissions at the field level for the consumed N or P in the food, driven by the household consumption of the food in the importing areas.

Potatoes rank as the fourth largest food crop worldwide, as well as in China [24,25]. They are planted widely in the six Chinese agricultural regions—the northeast (NE), northwest (NW), the north (NC), the middle and lower reaches of the Yangtze River (MLRYR), the southwest (SW), and the south of China (SC)—because their growth demands less water and fertilizer than cereal crops [26]. For a series of reasons, including mitigating the fertilizer increase and land competition, as well as the recommended diet for human health, the Chinese government proposed a potato staple food policy (PSF) in the year 2015 and set a target of 30% of the total potatoes to be consumed as a staple food in the future [27,28,29]. Per capita, potato consumption started to rise after the PSF from 2016 [29], and the harvested potato yield increased significantly in some regions of China [30]. Potatoes are widely cultivated throughout China with large differences in the new N and P utilization, the irrigation water consumption, the precipitation during the potato growing season, the net GHG emissions, and the yields in the six potato cultivation regions [12]. In addition, the commodity ratio of potatoes might be affected by different management practices, which impact potato size and quality [16]. Therefore, we hypothesize that there are differences in the VNNF, VNPF, VIWF, VTWF, and VCF between the six potato cultivation regions. With China’s PSF and the potential increase in the potato trade between different production regions, there is a need to calculate the VNNF, VNPF, VIWF, VTWF, and VCF indicators for each unit of potato consumed by a household. We further compare these five indicators with the VNF, VPF, irrigation water use efficiency (IWUE), WUE, and GHGI indicators, which are calculated from the potato field production perspective, to demonstrate the importance of the VNNF, VNPF, VIWF, VTWF, and VCF indicators from the perspective of food consumption. These consumption perspective indicators can not only help to accurately track the virtual new N and P, irrigation water, the total water inputs, and the total GHG emissions of the exported potatoes caused by the potato consumption of households in a given region, but they can also give some guidance for reducing the virtual resource inputs and GHG emissions of the imported potatoes from the supply–consumption chain.

2. Methodology

2.1. Description of the Study Boundaries

The study boundary of this study focuses on the abovementioned six potato production regions in China. In 2017, NE (Heilongjiang, Jilin, Liaoning), NC (Beijing, Tianjin, Hebei, Henan, Shandong, Shanxi), NW (Inner Mongolia, Shaanxi, Gansu, Ningxia, Qinghai, Xinjiang), MLRYR (Hubei, Hunan, Jiangxi, Anhui, Jiangsu, Zhejiang, Shanghai), SW (Sichuan, Chongqing, Yunnan, Guizhou, Tibet), and SE (Fujian, Guangdong, Guangxi, Hainan) accounted for 5.9%, 6.8%, 31.6%, 7.5%, 45.1% and 3.1% of the total potato cultivation area, respectively [24,31].

With the increasing frequency of food trade between different regions, many studies focus on analyzing the virtual N, P, water and land inputs, and GHG emissions of the imported or exported food between two regions [1,2,3,4]. Chemical fertilizers N, irrigation water N, atmospheric N deposition, and BNF were considered as the imported new N from outside the food system in many studies [8,20,32], which distinction with the N recycled inside the food system, such as organic manure, crop residues, food wastes, and sludge applied to fields [8,20,32]. The ratio of organic fertilizer application area to the total sown area of potatoes is relatively low. Meanwhile, the amounts of N and P from organic fertilizers were not directly reported as that of chemical fertilizers, and they are difficult to calculate because of the complex types of organic fertilizers and unavailable data on the contents of N and P in some organic fertilizers [27]. Therefore, we are placing emphasis on the new N and P inputs for potatoes at the field level. The new N input included N input from chemical fertilizers, irrigation water, atmospheric N deposition, and BNF [24]. The new P input included P from the first three sources of the new N plus P contained in the applied pesticides. Two indicators have been reported in our previous study [24]. Normally, water input for crops included irrigation water, precipitation, and the changes of soil water storage in a given depth before potato sowing and in harvesting time minus surface runoff and downward flux below the crop root zone [33]. Soil water storage variation, surface runoff, and downward flux below crop root zone are reported only in some sporadic studies. Besides, surface runoff can be negligible because there were separating strips that limited possible runoff along the furrows and downward flux below the crop root zone is very small relative to irrigation water and precipitation in dryland regions [33]. Therefore, water input in potato fields included only irrigation water and precipitation during the potato growing season in this study. We calculated virtual irrigation water and total water (irrigation water + precipitation) input per unit of the actually consumed potato standard yield. The GHG emissions included soil nitrous oxide (N₂O) and methane (CH₄) emissions, CO₂e emitted from the utilization of chemical N, P, and potassium (K) fertilizers, power used specifically for agricultural irrigation, fuel in farm operations, pesticides and plastic film used for mulching, and included annual soil organic carbon (SOC) sequestration [17,20,34]. The net GHG balance of each potato cultivation region was calculated under the conventional and optimized measures in our previous study [24] by considering soil N₂O and CH₄ emissions, annual SOC storage change (δSOC, kg C ha⁻¹ yr⁻¹), and the CO₂e fluxes from agricultural inputs in potato sowing to harvest using a life cycle approach (Equation (1)) [17,20,24].

Net GHG balance (kg CO₂e ha⁻¹) = (265.0 ×a + 28.0 × b + 8.3 × c + 1.5 × d + 1.0 × e
+ 1.3 × f + 2.6 × g + 18.1× h + 19.0 × i) − δSOC/12 × 44

(1)

where the lower-case letters represent the amounts of soil N₂O and CH₄ emissions and different agricultural inputs, which were cited from the reference [24]; the numerals 265 and 28 represent the CO₂e emissions per unit of soil N₂O and CH₄ emissions over a 100-yr scale, respectively [35]; and the other numerals represent the CO₂e emissions per unit of different agronomic inputs in potato production; these indicators were cited from the reference [20,24].

We found all potato cultivation regions acted as a net GHG source and showed large site dependency, varying from 2242.3 ± 737.1 kg CO₂e ha⁻¹ for NE to 3682.7 ± 1024.8 kg CO₂e ha⁻¹ for NC, and 836.8 ± 267.6 kg CO₂e ha⁻¹ for NE to 1824.9 ± 512.2 kg CO₂e ha⁻¹ for NC under local conventional and optimized measures, respectively [24]. Generally, the waste and loss of potatoes have no impact on GHG emissions during the field cultivation process. Nevertheless, waste and loss of potatoes between the available potato supplies and the actual consumption by a household will generate some impact on virtual GHG emission per unit actually consumed potato by a household. We, therefore, further calculated the VCF indicator of potatoes based on the abovementioned net GHG balance in our previous study [24], commodity ratio (Table S1), and the ratio between household available food items supplies and their actual consumption [14,15].

Two potato cultivation modes, namely conventional and optimized measures, were analyzed in this study. The conventional measure refers to local farmers’ experience management practices, which are mainly preoccupied with harvest yields, by applying relatively high water and fertilizer. The optimized practice refers to the knowledge-based measures aimed at improving potato yield by optimizing fertilizer N and water management practices and/or the cultivation modes and plant density of potatoes. We calculated the VNNF, VNPF, VIWF, and VTWF indicators of potatoes under conventional and optimized measures in the six cultivation regions from the perspective of consumption, based on the actually consumed potato or potatoes’ N by a household and the collected database of the new N and P, irrigation water and total water inputs in our previous study [24], combining with the newly collected commodity ratio of potato at the field level (Table S1) and the differences between available potato supply and its actual consumption [14]. Meanwhile, we calculated the VNF, VPF, IWUE, WUE, and GHGI indicators of potatoes under conventional and optimized measures in the six cultivation regions from the perspective of production, based on the harvested potato or potatoes’ N in the production system and the database of the new N and P, the irrigation water and total water inputs, and net GHG balance in our previous study [24]. The calculation boundary of this study is shown in Figure 1.

2.2. Data Collection

The data of the new N and P utilization, irrigation-water consumption, net GHG balance, and yield per unit area in the six potato cultivation regions under conventional and optimized measures have been reported in our previous study [24]. In addition, we collected the commodity ratio of potatoes at the field level and precipitation data during the potato growing season in the six potato cultivation regions under the two abovementioned management practices. Two indicators were directly collected from the peer-reviewed literature from 2000 to 2020 using ISI-Web of Science (Thomson Reuters) and the China Knowledge Resource Integrated (CNKI) database. In total, 206 results were collected for precipitation, which were divided between each of the potatoes cultivation regions as follows; NE (18), NC (29), NW (63), MLRYR (6), SW (73) and SC (17) (Table S2).

Potato cultivation is mainly driven by the final consumption of potatoes by households; at the same time, there is an underlying assumption that any potatoes that do not meet human consumption standards are thrown out or wasted in their harvest stage, even in many countries that use potato that is not considered fit for human consumption to feed animals. In other words, potatoes that are thrown as waste or used for feeding animals are the by-products of potato production for human consumption. Hence, the exclusion of potatoes that are thrown as waste or given to animals would not impact overall virtual nutrients and GHG accounting of the per unit actually consumed potato by a household. The proportion of potatoes that unmatch consumption standards to the total harvested potato is called the commodity ratio [14]. 152 and 145 commodity ratio results were collected for conventional and optimized measures, respectively, which were divided between each of the potato cultivation regions as follows; NE (22 and 21), NC (14 and 14), NW (45 and 43), MLRYR (27 and 27), SW (23 and 22) and SC (21 and 18). These values (Mean ± SE) fell into the range 75.2% ± 1.7% to 86.7% ± 2.0% and 81.6% ± 1.7% to 89.5% ± 1.6% for conventional and optimized measures, respectively (Table S1), close to the global mean food commodity ratio 80.0% for roots and tubers at the field level in mid-/high-income regions including the EU, North America, Oceania, and Industrialized Asia [14]. In addition, there are some food loss and waste during the storage, processing, distribution, and consumption stages [1,23]. In the calculations of virtual nutrient and GHG emission, we use the mean ratio of the available tuber (including mainly potatoes, sweet potatoes, yams, and tania) supply for humans to the actual tuber consumption by a household in China. The reported ratio of tubers supply to consumption was 1.2 to 1.7 in reference [15]; however, this ratio was underestimated by 5 times because they divided the standard yield of the available tuber supply by the fresh weight of tuber actually consumed by households. Hence, we recalculated the mean ratio was 7.7 ± 0.5 kg tuber supply (excluding tuber storage variation, net import of tuber and tuber for other uses) for per kg actual tuber consumption during the period from 2010 to 2015 based on the former research [24]. We use this parameter for potatoes in associate indicators calculation because potatoes’ yield accounts for more than 60.0% of the total tuber yield after 2016 [30], and there have unavailable specific parameters for potatoes.

2.3. Calculations of the VNF, VPF, IWUE, WUE, GHGI, VNNF, VNPF, VIWF, VTWF, and VCF of Potato

According to the abovementioned definition, VNF, VPF, IWUE, WUE, GHGI, VNNF, VNPF, VIWF, VTWF, and VCF can be calculated by Equations (2)–(11):

VNF (kg N kg⁻¹ N in the harvested potato) = (Chemical N fertilizers + Irrigation water N + Atmospheric N deposition + BNF)/(Yield × n)

(2)

VPF (kg P kg⁻¹ P in the harvested potato) = (Chemical P fertilizers + Irrigation water P + Atmospheric P deposition + Pesticides)/(Yield × p)

(3)

IWUE (m³ kg⁻¹ the harvested standard yield) = Irrigation water/Standard yield

(4)

WUE (m³ kg⁻¹ the harvested standard yield) = (Irrigation water + Precipitation)/Standard yield

(5)

GHGI (kg CO₂e kg⁻¹ the harvested standard yield) = Net GHG balance/Standard yield

(6)

$$\mathrm{VNNF} (\mathrm{kg} \mathrm{N} \mathrm{kg}{-}^1 \mathrm{N} \mathrm{in} \mathrm{the} \mathrm{consumed} \mathrm{potato})=(\mathrm{Chemical} \mathrm{N} \mathrm{fertilizers}+\mathrm{Irrigation} \mathrm{water} \mathrm{N}+\mathrm{Atmospheric} \mathrm{N} \mathrm{deposition}+\mathrm{BNF})/(\mathrm{Yield}\times {\mathrm{n}}_{ }\times \mathrm{R} \times \frac{1}{\mathrm{r}})$$

(7)

$$\mathrm{VNPF} (\mathrm{kg} \mathrm{P} \mathrm{kg}{-}^1 \mathrm{P} \mathrm{in} \mathrm{the} \mathrm{consumed} \mathrm{potato})=(\mathrm{Chemical} \mathrm{P} \mathrm{fertilizers}+\mathrm{Irrigation} \mathrm{water} \mathrm{P}+\mathrm{Atmospheric} \mathrm{P} \mathrm{deposition}+\mathrm{Pesticides})/(\mathrm{Yield}\times {\mathrm{p}}_{ }\times \mathrm{R}\times \frac{1}{\mathrm{r}})$$

(8)

$$\mathrm{VIWF} ({\mathrm{m}}^3 \mathrm{kg}{-}^1 \mathrm{the} \mathrm{consumed} \mathrm{standard} \mathrm{yield})=\mathrm{Irrigation} \mathrm{water}/({\mathrm{Yield}}_{ }\times \mathrm{R} \times \frac{1}{\mathrm{r}})$$

(9)

$$\mathrm{VTWF} ({\mathrm{m}}^3 \mathrm{kg}{-}^1 \mathrm{the} \mathrm{consumed} \mathrm{standard} \mathrm{yield})=(\mathrm{Irrigation} \mathrm{water}+\mathrm{Precipitation})/(\mathrm{Yield}\times \mathrm{R} \times \frac{1}{\mathrm{r}})$$

(10)

$$\mathrm{VCF} (\mathrm{kg} {\mathrm{CO}}_2\mathrm{e} \mathrm{kg}{-}^1 \mathrm{the} \mathrm{consumed} \mathrm{standard} \mathrm{yield})=\mathrm{Net} \mathrm{GHG} \mathrm{balance}/(\mathrm{Standard} {\mathrm{yield}}_{ }\times \mathrm{R} \times \frac{1}{\mathrm{r}})$$

(11)

where n and p represent N (0.30%) and P (0.14%) contents in potatoes [11,36], and R represents the commodity ratio of potatoes under conventional and optimized measures in China. We separately collected this parameter for conventional and optimized measures which have been directly reported in the publications (Table S1); r represents the aforesaid ratio of the available tuber supply to the actual tuber consumption by a household in China; For convenience to compare with other crops, we calculated IWUE, WUE, GHGI, VIWF, VTWF, VCF indicators based on standard yield. The standard yield represents fresh potato yield converted to the standard yield of grain by multiplying a conversion coefficient of 0.2 [37].

2.4. The Normalization Score of the VNNF, VNPF, VIWF or VTWF, and VCF Indicators in the Six Regions

To evaluate the potential difference between current potato plantation and ideal cultivation sites, we comprehensively assessed the VNNF, VNPF, VIWF or VTWF, and VCF indicators by summing the normalization scores of four indicators under conventional practice; the calculation principle was detailed described in our previous study [24]. We first set the maximum, and minimum scores of VNNF, VNPF, VIWF, and VCF as 0 and 1, respectively, and finally summed the normalization scores of the four indicators [38,39], as we did in the previous study [24]. Hence, the higher the normalization score, the lower the virtual new N and P, and irrigation-water inputs and GHG emissions, for produce per unit potato or potatoes’ N and P by household actual consumption, the more suitable for growing potato.

2.5. Uncertainty Analysis

The new N and P utilization, irrigation-water consumption, net GHG balance, and potato yields were directly cited from our previous study (Table S3) [24]. The uncertainties of these data were calculated by dividing the standard deviation by the corresponding mean value of each dataset. In this study, we also set uniform criteria for collecting the precipitation data during the potato growing season to calculate the total water input and the commodity ratio of potatoes under the conventional and optimized measures. The means and variation ranges of the precipitation and the commodity ratio of potatoes were calculated at the 90th confidence interval (Tables S1 and S2). In addition, the uncertainty was about 6.5% for the ratio of the available tuber supply to the actual tuber consumption, which was calculated by dividing the abovementioned standard deviation of the available tuber supply to the actual tuber consumption by households (0.5) by the mean ratio of the original tuber supply to the actual consumption by households (7.7) from 2010 to 2015, from reference [15,24]. The uncertainty of the VNNF indicator in the six potato cultivation regions was calculated using the error propagation equation from mathematical statistics [20,40]. First, we calculated the uncertainties of the new N and P inputs by the amount of chemical fertilizer, irrigation water, atmospheric deposition, BNF, or pesticides and the corresponding percentage uncertainties associated with them (Equation (12)). Then, we calculated the uncertainties of the VNF, VPF, VNNF, and VNPF indicators by the uncertainties of the total new N and P input, the yield, the N and P contents in potatoes, or the commodity ratio and the ratio of the available tuber supply to the actual tuber consumption by households in China (Equation (13)). Equation (13) was also used to calculate the uncertainty for the irrigation water plus precipitation amount.

$${\mathrm{U}}_{\mathrm{total}1}=\frac{\sqrt{{\left({\mathrm{U}}_1\times {\mathrm{x}}_1\right)}^2+{\left({\mathrm{U}}_2\times {\mathrm{x}}_2\right)}^2+\cdots +{\left({\mathrm{U}}_{\mathrm{n}}\times {\mathrm{x}}_{\mathrm{n}}\right)}^2}}{{\mathrm{x}}_1+{\mathrm{x}}_2+\cdots {\mathrm{x}}_{\mathrm{n}}}$$

(12)

$${\mathrm{U}}_{\mathrm{total}2}=\sqrt{{\mathrm{U}}_1^{\prime 2}+{\mathrm{U}}_2^{\prime 2}+\cdots +{\mathrm{U}}_{\mathrm{n}}^{\prime 2}}$$

(13)

where U_total1 and U_total2 are the combined uncorrelated uncertainty in the sum of the quantities and in the product of the quantities, respectively; x_i and U_i are the uncertain quantities and the corresponding percentage uncertainties associated with them [20,40]; U_i’ is the percentage uncertainties associated with different items in the calculations of IWUE, WUE, GHGI, VIWF, VTWF, and VCF indicators.

3. Results and Discussion

3.1. The VNF and VNNF of Potatoes in the Six Production Regions

The VNFs of potatoes at the field level showed large differences between the six potato cultivation regions (Figure 2a). The lowest VNF of potatoes was in the NE (1.8 ± 0.7 kg N kg⁻¹ N in the harvested potato), mainly due to the integrated impacts of a relatively lower chemical fertilizer input and a higher yield output in the NE than in other regions [24]. In contrast to the NE region, the highest VNFs of potatoes were 3.0 ± 1.3 kg N kg⁻¹ N in the harvested potato in the NC and 3.0 ± 1.2 kg N kg⁻¹ harvested potatoes’ N in the SC, which were mainly due to a 46.5–47.5% increase in the chemical fertilizer N inputs, with an 11.0–14.8% lower yield in the two regions, compared to the NE region. The VNF of potatoes could be reduced by 9.0–24.3% by optimizing the management practices (Figure 2a). The relatively larger reduction potential appeared in the MLRYR (24.3%), NC (21.4%), and NE (22.1%) regions, with the potential for a reduction in the chemical fertilizer N inputs of 19.5%, 13.2%, and 12.5%, while the yields could improve by 14.0%, 5.6%, and 13.2%, respectively, in three regions. These values indicate that the harvested potatoes had 1.8–3.0 and 1.4–2.7 kg of the new N input for production per unit N under the conventional and optimized measures, respectively, in the six agricultural regions. Our results were close to the VNF values of 1.8–4.0 and 1.1–2.3 kg N kg⁻¹ N in the harvested potato under the conventional and optimized measures, respectively, in [26] and significantly lower than the VNF values of 3.6–4.7 kg N kg⁻¹ starchy roots’ N based on the literature [9,10]. With similar amounts of the new N inputs (163.7–199.0 kg N ha⁻¹ in the literature vs. 177.5–264.2 kg N ha⁻¹ in the conventional measures in this study), the collected potato yields in previous studies from the statistical data of the National Bureau of Statistics of China [41] and the FAO database [25] were lower by 32.7–55.8% than our yield data collected from the conventional measures in the published literature. The values in China were somewhat higher than the 1.0–1.3 kg N kg⁻¹ N in the harvested potato in the USA and the EU [25,42] and the 0.9–1.4 kg N kg⁻¹ harvested potatoes’ N in Njoro, Kenya [43]. In addition, they were significantly higher than the 0.5–0.8 kg N kg⁻¹ N in the harvested potato in different water modes in southern Manitoba, Canada, in 2010–2011 [44].

The VNF indicator represents the new N input per unit of the harvested products’ N at the field level [14]. The virtual N input for the actual food consumption by households will be underestimated by the VNF indicators multiplied by the corresponding number of food items because this indicator overlooks the gap between the food supply and the actual consumption, which results in food loss and waste [14,15]. We, therefore, proposed the VNNF indicator to calculate the VNF per unit of the actual food N consumed by households. The VNNFs of the actual consumption of potatoes also showed a larger difference in the six potato cultivation regions, varying from 17.8 ± 7.8 to 30.1 ± 17.0 kg N kg⁻¹ N in the consumed potato in the NE and the MLRYR, respectively, under the conventional measures (Figure 2b). They could be reduced by 11.9–29.0% by improving the potato yield by decreasing the chemical fertilizer N by 12.5%, 19.5%, and 13.2% in the NE, NC, and MLRYR regions, respectively, combined with optimized water management practices and/or the cultivation modes and density of potatoes (Figure 2b). These values indicate that the potatoes consumed by households had 17.8–30.1 and 13.0–23.5 kg of the new N input for production per unit N under the conventional and optimized measures, respectively, in the six potato cultivation sites. The VNNF was 34.8 kg N kg⁻¹ N in the consumed starchy root for the year 2000, based on the reported new N input from the chemical N fertilizer, BNF, irrigation N plus atmospheric deposition N, starchy root yield, and the commodity ratio (80.0%) in reference [11], along with the ratio of the original tuber supply to the actual tuber consumption by households (7.7) reported by reference [24]. The mean ratio of tuber supply to consumption used in this study is significantly higher than the ratio of tuber supply to consumption 1.4 for tubers in China in reference [45]; however, whose calculation involves potatoes used for food, feed, and other uses, in addition, from supplemental information to see they overlooked the loss and waste rates of tubers during post-harvest stages and the consumption stage. A supply-to-consumption ratio of 7.7 implies that only 13% of tuber supply (this already excludes the production not matching consumption standards) is actually consumed as food. Close to 10% of potato supply as foodstuff (this accounts for 80% of the potato yield at the field level) was actually consumed as food in 2005; the remains (88%) are edible but discarded or the by-products of food production and mainly recycled for use as animal feed [36,46]. The abovementioned potatoes that are thrown as waste or used for feeding animals are the by-products of potato production for human consumption. Hence, the exclusion of potatoes that are thrown as waste or given to animals would not impact overall virtual nutrients and GHG accounting of the per unit actually consumed potato by a household.

The VNNF indicators were about 8.9–10.2 and 8.6–9.4 times higher than the VNF indicators under the conventional and optimized measures, respectively, in the six potato cultivation sites. This indicates that the virtual new N input of the actual consumption of potatoes would be significantly underestimated by 88.7–90.2% and 88.4–89.4% under the conventional and optimized measures if the commodity ratio of the potatoes at the field level and the ratio of the actual tuber consumption by households to the original tuber supply in China were omitted [15]. In other words, the reduction in the VNNF of potatoes should pay attention to improving the commodity ratio of potatoes and especially reducing the waste and loss of potatoes in the supply chain, in addition to the abovementioned measures to improve the yield.

3.2. The VPF and VNPF of Potatoes in the Six Production Regions

The VPFs of potatoes at the field level also showed large differences between the six potato cultivation regions under conventional measures (Figure 3a). The lowest VPF of potatoes was in the NE (0.9 ± 0.5 kg P kg⁻¹ N in the harvested potato), as it had the highest yield with a chemical fertilizer P input that was lower by 27.4–71.0% than the remaining five regions; in contrast, the highest VPF of potatoes was in the MLRYR region (1.8 ± 1.1 kg P kg⁻¹ P in the harvested potato) mainly caused by its use of the highest chemical fertilizer P input and the lowest yield [24]. These could be reduced to 0.6 ± 0.2 kg P kg⁻¹ harvested potatoes’ P for the NE and to 1.5 ± 0.7 kg P kg⁻¹ P in the harvested potato for the SW under the optimized measures (Figure 3a). These values indicate that for the harvested potatoes, there were 0.9–1.8 and 0.6–1.5 kg of new P input for the production per unit P under the conventional and optimized measures, respectively, in the six cultivation regions. We analyzed more than 90.0% of the new P inputs from chemical fertilizers in the six potato cultivation regions in a previous study [24]. The average VPF of Chinese potatoes was 1.1 kg chemical fertilizer P kg⁻¹ P in the harvested potato under the optimized measures [24], close to the values in this study. Our study values were relatively lower than the value of 1.3–2.1 kg chemical fertilizer P kg⁻¹ P in the harvested potato in Njoro, Kenya [43]; however, except for the NE, they were higher than the value of 1.0 kg chemical fertilizer P kg⁻¹ P in the harvested potato in the USA, which was calculated based on the total fertilizer P input for roots/tubers and the corresponding potato yield [25,42], as well as the 0.7–1.0 kg chemical fertilizer P kg⁻¹ P in the harvested potato under different water modes in southern Manitoba, Canada, in 2010–2011 [44]. They were also significantly higher than the value of −0.4 kg P kg⁻¹ P in the harvested potato in the EU [2].

As with the VNF and VNNF, we proposed the VNPF indicator for tracking the virtual P input for the potatoes’ P actually consumed by households. The VNPFs of the potatoes ranged from 8.4 ± 5.0 kg P kg⁻¹ P in the consumed potato for the NE to 18.8 ± 11.3 kg P kg⁻¹ P in the consumed potato for the MLRYR under the conventional measures (Figure 3b). These could be reduced by 10.5–31.5% by decreasing the fertilizer P by 15.9%, 16.3%, and 15.3% in the NE, NC, and MLRYR regions, respectively, integrated with the optimized N and water management practices and/or the cultivation modes and density of potatoes (Figure 3b). The VNPF values of our study were relatively higher in the P cost range of 5 to 13 kg P for delivering kg⁻¹ food P entering households for China as a whole from 1980 to 2005 [32]. The VNPF indicators would be underestimated in the same way as the abovementioned VNF and VNNF. This again shows the importance of taking into account the commodity ratio and the ratio of the actual tuber consumption by households to the original tuber supply in China in the calculations of the virtual resource input of the potatoes actually consumed by households, as well as for other crops, because of the loss and waste of many foods in industrialized Asia [14]; large gaps exist between food consumption and the corresponding supplies in China [15,24].

3.3. The IWUE, WUE, VIWF, and VTWF of Potatoes in the Six Cultivation Regions

Aside from the mainly rainfed potato fields in the MLRYR, NE, and SC regions, the IWUEs of potatoes varied from 0.03 ± 0.1 to 0.2 ± 0.2 m³ irrigation water kg⁻¹ standard yield at the field level under the conventional measures (Figure 4a). The WUEs of potatoes varied from 0.5 ± 0.2 to 1.0 ± 0.4 m³ total water kg⁻¹ standard yield at the field level, under the conventional measures, in the six cultivation regions. The IWUEs and WUEs could be improved to 0.02 ± 0.07 to 0.1 ± 0.1 m³ irrigation water kg⁻¹ standard yield and 0.5 ± 0.2 to 0.8 ± 0.4 m³ total water kg⁻¹ standard yield, respectively, under the optimized techniques (Figure 4b). The IWUEs in the NC and NW under the conventional practices fell into the range of the 0.05 m³ irrigation water kg⁻¹ standard yield for Manitoba, Canada [44] and the 0.3–0.4 m³ irrigation water kg⁻¹ standard yield under different irrigation modes in Njoro, Kenya [43]. The IWUEs in the NE, MLRYR, SW, and SC regions were significantly lower than the reported values in reference [44] due to the mainly rainfed potato fields in the four regions.

Generally, the IWUE and WUE are used to compare the irrigation water and total water use for different crops or for a given crop under different water management practices [16,47]. Nevertheless, the two indicators are unsuitable for calculating the water demand of the food actually consumed by households because they only represent the irrigation water or total water input per unit of the food harvested at the field level. The VIWF and VTWF indicators represent the virtual irrigation water and total water input for production per unit of the actual consumed standard yield at the regional scale in China, by further considering the commodity ratio of potatoes and the difference between the potato supply and its final consumption [18,19], in addition to the IWUE and WUE at field level. The VIWF and VTWF indicators are, therefore, better than the IWUE and WUE for modeling the water footprint of the actual food consumption by individuals or households.

Taking precipitation into account, the VTWF values ranged from 4.8 ± 2.2 m³ total water kg⁻¹ consumed standard yield to 9.3 ± 3.7 m³ total water kg⁻¹ consumed standard yield under the conventional management practices (Figure 4b). They could be reduced by 8.7–24.7% by optimizing water management practices. These values indicate that there is 4.8–9.3 and 4.3–7.9 m³ total water demand for production per unit of the potatoes consumed by households, under the conventional and optimized measures, respectively, in the six cultivation regions. The VIWF and VTWF indicators are also 8.9–10.2 and 8.6–9.4 times higher than the IWUE and WUE indicators, under the conventional and optimized measures, among the six potato production regions. This indicates that the virtual irrigation and total water inputs of the final consumed potatoes would be significantly underestimated by 88.7–90.2% and 88.4–89.4%, under the conventional and optimized measures, respectively, if the commodity ratio of the potatoes at the field level and the ratio of the actual tuber consumption by households to the original tuber supply in China were omitted.

3.4. The GHGI and VCF of Potatoes in the Six Agricultural Regions

The GHGI indicator provides a platform for comparing the overall effects of different cropping systems on the GHG emissions per kilogram of yield at the field level [16,17,18]. The GHGI of potato ranged from 0.3 ± 0.1 kg CO₂e kg⁻¹ standard yield for the NE to 0.6 ± 0.2 kg CO₂e kg⁻¹ standard yield for the NC, under the conventional measures (Figure 5a), close to the value of 0.6 ± 0.3 kg CO₂e kg⁻¹ standard yield for the Chinese potato as a whole, and these values fell into the range of the 0.2–2.3 kg CO₂e kg⁻¹ for the major crops in China [20], except for the NE region. The lowest GHGI of potato was in the NE, mainly caused by it obtaining the highest yield with relatively low chemical fertilizers’ N input and almost no power used for irrigation due to the rainfed conditions. This was significantly higher than the 0.1 kg CO₂e kg⁻¹ standard yield for potatoes in Iran [48] because of the high fertilizer input and relatively low yield of Chinese potatoes. This could be reduced by 47.1–67.6% by increasing the SOC storage and decreasing the abovementioned chemical fertilizer N and P inputs in the NE, NC, and MLRYR regions, as well as decreasing the power used for irrigation in the NC, under the optimized measures [24].

The GHGI indicator is unsuitable for direct calculation of the GHG emissions per unit of the foods consumed in the receiving regions because it fails to consider the commodity ratio at the field level and the loss and waste of foods in the food supply chain [12,14]. We, therefore, proposed the VCF concept and analyzed the VCFs of potatoes in the six cultivation regions. The VCFs ranged from 3.4 ± 1.5 to 5.9 ± 2.4 kg CO₂e kg⁻¹ consumed standard yield under the conventional measures (Figure 5b). This could be reduced by 48.8–69.7% by closing the large gaps between the available potato supply and the actual consumption by households [19], apart from optimizing the management practices for GHGI reduction at the field level. Our results indicate that the associated GHG emissions in the potato production area, caused by the actual potato consumption of a given region, were underestimated by 88.7–90.2% and 88.4–89.4%, under the conventional and optimized measures, respectively, when the estimation was based only on the traditional GHGI perspective, which neglected the commodity ratio at the field level and the gaps between the originally harvested products and the actual consumption [14,15].

Our results highlight the significance of studying the regional-scale VNNF, VNPF, VIWF, VTWF, and VCF indicators per unit of potato consumed in different cultivation regions from a consumption perspective. The environmental assessment model of the actual potato consumption by households can directly use these improved indicators. Our results also highlight that the VNNF, VNPF, VIWF, VTWF, and VCF indicators per unit of potato consumed have a large potential for reduction by improving the commodity ratio of potatoes and especially reducing the loss and waste of potatoes in the supply–consumption chain, in addition to increasing the potato yield by optimizing the fertilizer N and P inputs in the NE, NC, and MLRYR regions and optimizing the water management practices and/or the cultivation modes and plant density of potatoes at the field level. Similar research is needed for other food crops because the underestimation of the GHG emissions and the abovementioned new N, P, and water inputs for the potatoes actually consumed by households could occur in other crops as well.

3.5. Comparison of the Integrated Production Efficiency of the Six Potato Cultivation Regions

We found the NE, SW, and NW regions were relatively more suitable for planting potatoes than the three remaining regions based on the normalization score of the new N and P use efficiencies and the GHGI indicators in the previous study [24]. The results were consistent, excluding the VIWF (Figure 6a) and VTWF (Figure 6b) indicators in this study. Taking the irrigation water and total water (irrigation water + precipitation) into account, the normalization score in the SC region was significantly improved due to the relatively higher WUE (Figure 6). The NE remains the most suitable region for potato cultivation, based on the VNNF, VNPF, VIWF or VTWF, and VCF indicators from the perspective of household consumption, followed by the SW, SC, and NW regions. Our results agree with the Chinese government’s measures to accelerate potato production, increase per capita potato consumption and promote the PSF, including increasing potato cultivation after harvesting the late rice crop in the SC region and planting more potatoes by converting a small proportion of the maize cultivation area to potato in the “Sickle Bay” region (NE, NW, and SW) in China [24,28].

4. Conclusions

We comprehensively analyzed the VNNF, VNPF, VIWF, VTWF, and VCF indicators of the potatoes consumed by households in the six Chinese potato cultivation regions, by considering the harvested potato yield, the commodity ratio of potatoes at the field level, and the ratio of the potatoes consumed to the available potato supply. These indicators showed large differences between the six potato cultivation regions caused by different fertilizer inputs, water management, yields, and the commodity ratio of the harvested potato. The NE is the most suitable region for cultivating potatoes, followed by the SW, SC, and NW regions, based on the normalization score of the VNNF, VNPF, VIWF or VTWF, and VCF indicators from a consumption perspective. Moreover, a large potential exists to improve the VNNF, VNPF, VIWF, VTWF, and VCF indicators of potatoes, as we found they could be reduced by 11.9–29.0%, 10.5–31.5%, 25.6–35.8%, 8.7–24.7%, and 48.8–69.7%, respectively, compared to local farmers’ conventional practices. The VNNF, VNPF, VIWF, VTWF, and VCF indicators of potatoes were underestimated by 88.7–90.2% and 88.4–89.4%, under the conventional and optimized measures, respectively, when the calculations of these indicators stopped at crop harvest and omitted the commodity ratio at the field level and the ratio of the available potato supply to the potatoes consumed. Hence, the improvement measures for the VNNF, VNPF, VIWF, VTWF, and VCF indicators of potatoes should pay attention to closing the gaps between the available potato supply and its actual consumption by reducing potato loss and waste in the supply chain and improving potato yield and its commodity ratio by optimizing the fertilizer N input in the NE, NC, and MLRYR regions and the water management practices and/or the cultivation modes and density of potatoes at the field level. Our findings contribute to improving the existing policies for the sustainability of agriculture and climate from the whole field production–supply–consumption chain. With the increasing volume and frequency of the food trade, similar work is urgently needed for other crops because the differences in the virtual resource inputs and environmental footprints might occur in other Chinese crops under different soil–climatic conditions and management practices in different production regions.